search for: usereductionintrinsic

...bly proven themselves to be stable. However the error you're > seeing is because the AArch64 backend still expects to deal with only > intrinsics it can *natively* support, and i1 is not a natively supported > type for reductions. See the code in > AArch64TargetTransformInfo.cpp:useReductionIntrinsic() for where we > decide which reduction types we can support. > > For these cases, we need to implement more generic legalization support > in order to either promote to a legal type, or in cases where the target > cannot support it as a native operation at all, to expand it to a...

...> you're > > seeing is because the AArch64 backend still expects to deal with only > > intrinsics it can *natively* support, and i1 is not a natively > supported > > type for reductions. See the code in > > AArch64TargetTransformInfo.cpp:useReductionIntrinsic() for where we > > decide which reduction types we can support. > > > > For these cases, we need to implement more generic legalization > support > > in order to either promote to a legal type, or in cases where the > target > > c...

...that are exposed by other targets. Transitional API ================ To support both the old-style tree reduction patterns and the new intrinsics for clients, we can use a generalization of the TTI call we introduce in our SVE patches. We could provide something like: bool TargetTransformInfo::useReductionIntrinsic(unsigned Opcode, Type *Ty, bool IsMaxOp, bool IsSigned, bool NoNaN) In the prototype patch, I've created two separate ways to generate reductions from an IR generator's perspective, depending on...